This invention relates to devices used in and methods for making integrated circuits, and more particularly to high K dielectrics used in making integrated circuits.
Silicon dioxide has been by far the most common and effective insulator used in making integrated circuits. This has a very high level of integrity and, in particular, is able to be made with a very low defect density. The result is that the silicon dioxide operates very effectively in having low leakage. With regard to gate dielectrics, one of the desirable features of the dielectric is that it couple the overlying gate to the underlying channel so that the channel is responsive to the stimulus applied to the gate. In this regard it is desirable for that dielectric to have a high dielectric constant commonly known as K.
Currently there is much work being done in developing high K dielectrics that have a higher dielectric constant than that of silicon oxide. There are a number of those, but one of the advantages of silicon oxide is its high band gap which results in it being a very effective insulator. Thus, many of the materials being developed for high K purposes have been found to have problems because they do not have a high enough band gap or because they are difficult to make with enough integrity to prevent current leakage through the dielectric.
One of the characteristics that is desirable for the high K dielectric is that it be amorphous. It must remain amorphous for its entire life including during manufacturing and subsequently during functional operation as part of the completed integrated circuit. Many of the high K dielectrics have sufficiently high K and sufficient integrity at time of deposition, but over subsequent processing steps and the heating that is associated with that, the result is crystallizing of these films. These films that are so crystallized are not perfectly crystallized throughout their entire length and width but have areas known as grain boundaries between the crystalline structures that are formed. These grain boundaries are areas of leakage and other problems that affect electrical performance.
An alternative to amorphous is monocrystalline films. In theory, these films can be made typically monocrystalline. There are several problems with that. One is matching the crystalline structure of the film with that of the underlying semiconductor, typically silicon, as well as during the formation process that it be in fact perfectly formed. Epitaxial layers, that is layers that are monocrystalline, are known in the industry. Silicon can be made epitaxially. These epitaxial processes generally are relatively slow compared to other deposition processes. One of the techniques by which very small films can be put down in a monocrystalline form is molecular beam epitaxy. There are problems with this approach in that it is very slow so that the throughput, the number of wafers per a period of time, is very low compared to conventional deposition processes such as CVD. Thus, molecular beam epitaxy (MBE) is generally considered not a manufacturable technology. Even with using MBE technology there is still the difficulty of insuring defect free films. In order to achieve this, pressures must be extremely low and the process is very slow. One very thin layer, by thin meaning 10 to 30 angstroms, can easily take 2 hours on an MBE machine.
In developing new high K dielectrics there is also another potential problem of having too high of a dielectric constant. If the dielectric constant is too high, there is an effect that is called fringing field effect that adversely affects the performance of the transistor. This has to do with excessive coupling between the gate and the source/drain. Thus, the materials that are being developed desirably have a range typically between 20 and 40 for the dielectric constant. This range may change somewhat as the technology develops further.
Another aspect of a desirable high K dielectric is in terms of its equivalent capacitance to that of a certain thickness of silicon oxide. Silicon oxide has been so commonly and effectively used that it has become a standard and the industry often describes certain characteristics in terms of its relationship to silicon oxide. In this case, the typical desirable silicon oxide equivalent is between 5 and 15 Angstroms but with silicon oxide of 5 to 15 angstroms it has problems with leakage, reliability and growth rate. Thus, when a film is that small there can be difficulties in manufacturing it as well as using it. The desirable coupling is to have a dielectric that has the equivalence of the thickness of 5 to 15 angstroms of silicon oxide but a greater actual thickness. The actual minimum thickness that is generally believed to be desirable is about 25 Angstroms.
Thus, there is a need for a dielectric film which has a dielectric constant within a desirable range, the ability to be made of high integrity, a thickness in a desirable range, and the ability to be made in a manufacturing process.